Breaking through the automation glass ceiling with the Resolution Learning Loop™

It’s almost 2026. Can we all agree that AI is proving its ability to help automate service?

Sure, some organizations are still working through AI Agent proof-of-concepts or manually using ChatGPT to rewrite old Help Center articles.

But Zendesk customers—more than 20,000 of them—are using our AI platform to achieve amazing results. SeatGeek automated over half of all support interactions with AI agents, more than doubling satisfaction scores at the same time. And customers like Aeroport de Lyon—who reached an 85% automation rate—prove that 80% automated resolution rates aren’t a pipe-dream. They’re achievable. But not by magic.

Organizations seeing climbing automation rates all have one thing in common. They aren’t deploying AI and hoping for the best. They’re using AI to attack common problems and then—crucially—learning, iterating, and improving.

Continuous improvement with the Resolution Learning Loop™

This is exactly what Zendesk’s Resolution Learning Loop™ delivers. By analyzing all of your interaction data, Zendesk AI continuously improves your ability to deliver more automated, quality resolutions. That means your automated resolution rate today becomes your baseline tomorrow—and can only get better.

But the Resolution Learning Loop™ isn’t the standard flywheel of improvement you’d expect from any AI platform. What makes it possible–and unique–is Zendesk’s unified Resolution Platform. When your AI and human agents work together on the same platform that also includes all of your service data, knowledge, and tools, you can drive more improvement, faster.

Every interaction, workflow, and resolution makes the next one faster, smarter, and more reliable. With the Zendesk Resolution Learning Loop™, your automated resolution today is your baseline. Every time you talk to a customer is an opportunity to get better and better.

Here’s how.

Every service interaction begins somewhere: a Help Center visit, a chat with an AI agent, or an email. Whether it sparks connection or introduces friction, each is an opportunity to improve.

There’s gold in interaction details: tickets, conversations, agent actions, even data from external systems. Behind first response time (FRT), customer satisfaction score (CSAT), and average handle time (AHT) metrics, is the rich data where you’ll find opportunities to improve: the mistakes, SLA breaches, poor training, churn risks, and broken workflows. Before AI, most of this was hidden or only available to analysts after months of research.

This is where Zendesk AI shines.

We help you analyze and interpret all this data—spotting opportunities to improve agent performance, quality, knowledge, and workflows. Then we enable anyone to deploy smart fixes with Copilots and Builders, no code required. AI becomes the engine behind smarter interactions and stronger resolutions.

Every interaction gets smoother and you can make progress toward your automated resolution goals.

Let me give you an example of how this would work in real life.

Streamlining warranty claims

Let’s say you’re a manufacturer using Zendesk for service and your real-time monitoring reports alert you to a spike in warranty claims for a specific equipment SKU that’s impacting agent productivity and time-to-resolution. It’s unfamiliar and unplanned so your AI agents are escalating to humans, slowing everything down.

This is where the Resolution Learning Loop™ kicks in:

• Zendesk AI (QA, Analytics, Copilots) helps you verify the cause, scope, and solution by analyzing tickets, conversations, and the interaction with your ERP.
• Admin Copilot recommends an autoreply specific to the equipment SKU to immediately reduce the burden on human agents.
• Action Builder lets you set up and automate a copilot-assisted claims workflow that verifies warranty terms and availability of the replacement, authorizes the shipment, and notifies the account team via Slack.
• App Builder’s natural language prompts let you create a side bar app in minutes that pulls product and warranty info right into agents’ workspace.
• Knowledge Builder leverages generative AI to turn insights from warranty conversations into an alert on your help center to guide customers.

Now your AI Agents can automate this entire process, with humans approving the final replacement step if you want.

With improved knowledge and automations, customers get what they need, agents are more effective, and time to resolution improves. Every future warranty issue can be triaged and fixed faster.

You’ve broken through the automation glass ceiling.

But now a warning.

Here’s what blocks the Resolution Learning Loop™:

• Siloed data. Interaction data sitting in bolted-on external AI agents or single-channel systems can’t fuel improvement. You’d need to integrate it with the rest of your data, and then apply AI. Slower, more expensive, more prone to breakage.
• Siloed tools. If AI agents aren’t connected to the same tools as human agents, they’ll hit a ceiling. If their interactions aren’t scored by QA, or they can’t access workflow tools, resolution quality and automation will stall.
• Manual, limited QA: Without AI-driven QA that goes beyond scoring and coaching to analyze conversation data, you’ll miss insights to improve processes and fill knowledge gaps.
• Inertia, or fear. Testing and POCs are good, but competitors aren’t standing still. If you’re not careful, you’ll be watching their backs as they win the race.

Imagine removing these blockers and automating just 10% more so your agents can better serve top customers.

We’re here to help

In 2025, 20.7% of the world population will have used Zendesk to get service from companies, their employer, or their government. We’ve already helped 20,000 organizations on their journey to agentic service. And we’ve got the blueprint to help any organization take the next step, no matter what stage you’re at.

To learn more, contact your Zendesk representative or account team.